I’m using chatgpt assistants API (retrieval). I have uploaded a file and passed the fileId to the thread and then if I run the thread, it is not considering the entire JSON file while answering the questions.
Also which is the best file format to upload and ask answers from. Uploading CSV file with assistants api for retrieval is not working.
If I have a JSON file, the key values can be different from what the user may ask from because the JSON file contains data of a table. let’s say in the JSON, it will have orderName key and I will show as ‘Order Name’ in the table. so user will ask questions based on the Order Name, so how can I ask the chatgpt api to recognize it. Is there a way by which I can better do this?
Hi there - the retrieval tool is best used when it makes sense to “search” the file for information. If you’d always like the full contents of the file to be used by the assistant, you should put the full file in the context of the model (e.g. you can create a message with the full content of the json file).
Hey Nikunj, just came across your answer regarding this.
I’ve been toying around with this process, trying to get it to effectively sweep an entire file, but what you’re saying makes sense.
Are you talking about taking the file contents and sticking it completely within the instruction. Or are you talking about creating a message chain within the thread IE: First message would be here is the data set soon to follow the “customers prompt” so to speak?
If I’ve got anything wrong, or I’m completely off base please correct me. Much appreciated!